skottenborg

Yes! Or even better, use Proxmox snapshots during upgrades for easy rollback on failure

Though it may be somewhat annoying, I think forcing named parameters would be a big win for clarity. Is this something you have considered?

Internettet er vist ikke så stort :-)

Please elaborate.

I immediately thought this sounded like an LLM reply.

The people behind it are quite popular SWE YouTubers, so that might explain the reach :-)

Great work, I like your website design! But I don't think you should fool people by writing "Join thousands of students studying with LIRA." unless that really is the case (which I doubt[0]).

I'm building some different, but also in EdTech. What did you use for capturing the video demos?

[0]: I'm skeptical of the claim since the site is brand new, doesn't show up on Google and launched on PH today.

I don't know, but shilling it here is definitely not the place.

Interesting. Stephen Wolfram also wears a camera that snaps pictures every 20 seconds for special events.

This is quite a read: https://writings.stephenwolfram.com/2019/02/seeking-the-prod...

"(I work at OpenAI.)"

Ah yes, also known as being co-founder :)

It's a complete ecosystem with design and functionality being consistent and on par.

How did you figure out the UPS was the cause?

The internal DNS records are very handy if you host local services.

I doubt personal passions would merit the company funding required for such big models.

I'm not sure HTTPS is an option for this as it requires accurate time/date for encryption. Maybe the RTC could be valid, but I'm not sure it handles different time zones well and it might have lost time anyways.

Given this, it's interesting that an established company like Meta releases open source models. Just the other day Zuck mentioned an upcoming open source model being trained with a tremendous amount of GPU-power.

I could see a future where this works really well. It doesn't seem to be the case right now though.

The "super resolution" showcased in the video seemed almost identical to adjusting the "sharpness" in any basic photo editing software. That is to say, perceived sharpness goes up, but actual conveyed details stays identical.

Cool! Can webLLM handle inference of models with any meaningful size?

Can I ask what model is used?

Interesting. I think text chats would be a nice addition as they allow asynchronous conversations which would be beneficial.

You can also add internal DNS records with Pi-Hole if you happen to be self hosting that as well.

No, you can just create guest link for single use.

I can recommend PicoShare which I've been self-hosting for a while. It's a couple of years old now and still maintained. Also, it's very simple and allows for guests too.

https://github.com/mtlynch/picoshare

On creating EUV:

"In our laser-produced plasma (LPP) source, molten tin droplets of around 25 microns in diameter are ejected from a generator at 70 meters per second. As they fall, the droplets are hit first by a low-intensity laser pulse that flattens them into a pancake shape. Then a more powerful laser pulse vaporizes the flattened droplet to create a plasma that emits EUV light. To produce enough light to manufacture microchips, this process is repeated 50,000 times every second."

In my experience .NET core works great on Linux.

It has actually been fantastic for writing small jobs that run in Linux docker containers.

This is correct. It's golang and it exposes a plentiful API.

Out of curiosity, what use case can possibly warrant using Windows 7 in 2023?

Interesting. I have seen public decks of Leetcode questions that had a summary of the solution for each problem, but I've never heard of this take.

I'm tempted to try it for my self. Did you only include problems that you solved previously?

Would you care to expand on how Anki helped you with Leetcode problems? Did you just think of a solution in your head and then compared it with the back of the card?

I think the reason Microsoft was concerned about the new chatGPT release was due to the fact that you could prompt the model to a download link of the training files. Thus, if an employee trained a custom GPT on sensitive material you could quite easily retrieve the data.

It's very interesting how different reactions to caffeine are. I can take a 200 mg caffeine pill 5 hours before sleep and have no trouble sleeping. My Samsung smartwatch doesn't indicate any loss of deep sleep either (1-2 hours usually).

I do have a high tolerance though, but I don't weight much either.